Improvement in Endocardia1 Border Delineation Using Tissue Harmonic Imaging HARALD BECHER, M.D.," KLAUS TIEMA", M.D.," THOMAS SCHLOSSER," CHRISTOPH POHL, M.D.," NAVIN C. NANDA, M.D.,? MICHALAKIS A. AVERKIOU, PH.D.,$ JEFF POWERS, PH.D.,$ and BERNDT LUDERITZ, M.D.* *Department of Cardiology, University of Bonn, Bonn, Germany, ?Division of Cardiovascular Diseases, University of Alabama at Birmingham, Birmingham, Alabama; and tATL Ultrasound, Inc., Bothell, Washington

Background and Methods: For years, tissue has been assumed to be a linear medium in diagnos- tic ultrasound applications; thus, no backscattered signals in the second harmonic band are expected in harmonic imaging without the injection of a contrast agent. However, it has been shown that a useful tissue image is formed even without a contrast agent. The a im of this study was to evaluate whether this tissue harmonic image provided improved visualization of endocardial borders. Fifty- six adult patients with various heart diseases were investigated using conventional two-dimensional echocardiography and tissue harmonic imaging. I n 30 of these patients, the left ventricular endo- cardial borders were well defined in the standard parasternal and apical views using conventional two-dimensional echocardiography. I n the remaining 26patients, delineation of endocardial borders was not possible in at least two segments. The equipment used was an ATL HDI-3000 diagnostic sys- tem equipped with harmonic imaging. Results: I n all 56patients, the myocardium and valves could be imaged with tissue harmonic imaging. Harmonic recordings were sharper and contained fewer clutter artifacts than conventional recordings. Most striking was the enhancement of left ventricular endocardial borders. I n the 26 patients with incomplete delineation of left ventricular endocardial borders, wall motion could be evaluated in 290 of 312 (93%) segments with tissue harmonic imag- ing compared with only 168 of 312 (54%) segments with conventional echocardiography (P < 0.001). Conclusions: Tissue harmonic imaging improves image quality and can be used to enhance the def- inition of left ventricular endocardial borders. These findings can be explained by the nonlinear propagation of ultrasound within the tissue, which results i n distortion of the transmitted signal and, thus, harmonic generation. (ECHOCARDIOGRAPHK Volume 15, July 1998)

tissue harmonic imaging, left ventricular endocardial border delineation, clutter, artifacts

Correct delineation of endocardial borders is the prerequisite for confident assessment of left ventricular function. Despite tremendous technical advancement in echocardiographic instrumentation, there remains a substantial number of patients who cannot be adequately imaged with conventional two-dimensional

Address for correspondence and reprints: Priv. Doz. Dr. Med. Harald Becher, Department of Cardiology, Univer- sity of Bonn, Sigmund-Freud Str. 25, 53105 Bonn, Ger- many. Fax: 49-2281287-4323,

echocardiography. Contrast echocardiography may be an alternative procedure in some of these patients, especially when used in con- junction with harmonic imaging.' Harmonic imaging, the method in which an image is formed from the second harmonic component of the backscattered signal, originally was in- vented to use the nonlinear properties of con- trast microbubbles.' Contrast microbubbles resonate when exposed to diagnostic ultra- sound. Resonant microbubbles scatter not only ultrasound at the sonification frequency but

Vol. 15, No. 5, 1998 ECHOCARDIOGRAF'HY: A Jrnl. of CV Ultrasound & Allied Tech. 511

BECHER, ET AL.

512 ECHOCARDIOGRAPHY: A Jrnl. of CV Ultrasound & Allied Tech. Vol. 15, No. 5,1998

also harmonic frequencies, which are multiples of the transmit f req~ency.~ The spectrum of the backscattered ultrasound received by the transducer from the microbubbles consists of both the fundamental frequency and the har- monic frequencies. If in the receive phase only the second harmonic frequency is processed, signals from microbubbles would be selectively displayed, and those from tissue and blood would be suppressedP

If tissue is assumed to be a linear medium, no image would be formed if harmonic imaging is used in patients without the injection of con- trast. However, we were surprised to find good- quality images from valvular and myocardial structures during noncontrast harmonic imag- ing when the receiver gain was increased and optimi~ed.~ The current study was performed to evaluate this effect in consecutive patients and to assess its potential in the clinical setting.

Methods

Study Patients

The study population consisted of 56 consec- utive patients (mean age, 53 years; age range, 23-76 years; 42 men and 14 women) who were referred for an echocardiographic investigation before left heart catheterization. The indica- tion was suspected ischemic heart disease or valvular heart disease.

Ultrasound Image Acquisition and Analysis

Standard parasternal and apical views of the left ventricle were recorded with a diagnostic ultrasound system equipped with harmonic imaging (HDI-3000, ATL, Bothell, WA, USA). The output power was set to the highest possi- ble level (mechanical index = 1.5) when avail- able. Fundamental images were obtained at 2.7-MHz frequency and harmonic images at 1.67 MHz and 3.33 MHz receive frequency. Gain settings and transducer position were in- dividually optimized in each patient using the fundamental mode. At the end of a standard echocardiographic study using fundamental B- scan, the apical four-chamber view was recorded in the fundamental mode, and the imaging mode was switched to harmonic B-

scan, holding the transducer in the same posi- tion as in the fundamental B-scan. In this po- sition, the initial recordings were obtained with the same transmit and receive gain as in fundamental imaging. Subsequently, the re- ceiver gain was increased and individually op- timized as in fundamental B-scan to optimally visualize myocardial and valvular structures. The same procedure was performed using the apical two-chamber view.

In ten patients, additional recordings in fun- damental and harmonic modes were obtained using the HDI 3000 system with a different transducer with a higher transmit frequency of 2.3 MHz and a higher corresponding harmonic frequency of 4.6 MHz. All images were stored on S-VHS videotape using a commercially available videotape recorder as well as on a magneto-optical disk using the storage system of the ultrasound devices.

Two experienced observers, who were blinded to each other’s results, evaluated the recordings stored on the magneto-optical disc. Endocardia1 border delineation was scored in each of 12 segments using a 3-point scoring scale in which 0 indicates no delineation, 1 in- dicates poor delineation, and 2 indicates excel- lent delineation. A consensus of these two ex- perienced observers was required to use the segment score for further comparative analy- sis. Mean endocardial border delineation score was calculated for each segment in both funda- mental and harmonic modes. Using the area-length method, left ventricular volumes and ejection fraction were calculated from the stored frames after manual tracing of the en- docardial borders by one obse r~e r .~

Left ventricular volumes and ejection frac- tions were determined in all patients in whom the endocardial border delineation scores of all segments in the apical four- and two-chamber views were 2. These data were compared with the findings of left ventricular angiography, which was performed within 24 hours of the echocardiographic examination. Paired Stu- dent’s t-tests were performed to compare endo- cardial border delineation scores, left ventricu- lar volumes, and ejection fractions obtained from fundamental recordings with measure- ments from harmonic B-scans. The P value,

TISSUE HARMONIC IMAGING

mean difference, and limits of agreement were calculated.

Results

In 30 of 56 patients, fundamental imaging provided adequate recordings of the two apical views with good delineation of left ventricular endocardial borders. Adequate recordings were images that were classified as having an endo- cardial border delineation score of 2 by both readers in all segments. The recordings of the remaining 26 patients were judged as subopti- mal echocardiograms. With harmonic imaging and optimization of the receiver gain, endocar- dial borders, myocardial tissue, and valvular structures appeared enhanced and more echo dense compared with use of the fundamental mode (Fig. 1). In addition, spontaneous echo contrast within the left ventricle was found in 51 of 56 patients when the receiver gain was adjusted for best visualization of endocardial borders. All 30 patients with an endocardial border delineation score of 2 on fundamental imaging retained the same score in harmonic imaging. In these patients, left ventricular end-diastolic and end-systolic volumes, as well as ejection fractions, obtained with harmonic imaging were not significantly different from the values measured in the fundamental mode.

In the 26 patients with endocardial border delineation scores of < 2 in at least two seg- ments, wall motion could be evaluated in 168 of 312 (54%) segments with conventional echo- cardiography and in 290 of 312 (93%) segments with tissue harmonic imaging (P < 0.001) (Fig. 2). The improvement in left ventricular border delineation was statistically significant in the lateral segments (apical four-chamber view) and the anterior segments (apical two-cham- ber view) (Figs. 3 and 4). In comparison with fundamental mode, harmonic endocardial bor- der delineation scores were not significantly higher in septa1 and inferior segments, but the

also resulted in a better visual display of car- diac structures than did fundamental imaging. Spatial resolution was improved with higher frequencies (Fig. 5). Both lower and higher fre- quency transducers resulted in reduction in clutter artifacts in the harmonic mode com- pared with fundamental imaging.

Discussion

Difficult-to-image patients include individu- als who have thick chest walls, emphysema, or narrow intercostal spaces or who have had pre- vious surgery (coronary bypass surgery, breast implants, and so on). Many of these technically difficult patients undergo echocardiographic

Of these segments was less noisy and Figure 1. Corresponding frames of an apical two- chamber view obtained with fundamental (top) and

delineated using fundamental mode due to “noise.” With harmonic imaging, the lateral wall endo- cardium is fully displayed.

more pronounced.

ink? was Performed With a higher frequency transducer and a correspondingly higher fre- quency harmonic imaging, harmonic imaging

In the ten patients in whom imag- harmonic (bottom) imaging. The lateral wall is not

Vol. 15, No. 5 , 1998 ECHOCARDIOGRAPHY: A Jrnl. of CV Ultrasound & Allied Tech. 513

BECHER, ET AL.

Figure 4. Apical two-chamber view. Left ventricu- lar endocardial border delineation scores are shown in the same manner as in Fig. 3. An improvement of endocardial border delineation values was found in segments 4 and 5.

Figure 2. Left ventricular endocardial border de- lineation showing a comparison of fundamental and harmonic modes. The total numbers of segments with different endocardial border delineation scores are shown,

evaluation because of suspected ischemic heart disease. Left ventricular wall motion analysis by two-dimensional echocardiography is a key step in the noninvasive diagnostic workup of these patients.6 Thus, an improvement in imaging techniques would be expected to have a significant impact on clinical decision mak- ing in the management of these patients. The results of our preliminary study suggest that harmonic imaging without contrast is a signif- icant advance toward more confident evalua- tion of patients with suboptimal echocardio- grams with fundamental imaging. Our data clearly demonstrate the superiority of har- monic imaging without contrast over funda-

Figure 3. Apical four-chamber view. Left ventricular endocardial border delineation scores (mean values and standard deviation). (Top) Results of fundamental imaging. (Bottom) Results of harmonic imaging. An improvement in endocardial border delineation values was found in segments 10 and 11.

Figure 5. Corresponding frames of a five-chamber view obtained with a 2.3 14.6-MHz transducer. Left ventricular endocardial borders are obscured on the fundamental image (top) but are well defined on the harmonic recording (bottom).

514 ECHOCARDIOGRAPHY. A Jml. of CV Ultrasound & Allied Tech. Vol. 15, No. 5 , 1998

TISSUE HARMONIC IMAGING

Figure 6. Apical four-chamber view. The clot (ar- rowheads) in the left ventricular apex is much better visualized using tissue harmonic imaging (A) and power imaging (B) compared with fundamental imaging (C). L A = left atrium; RA = right atrium; RV = right ventricle.

mental echocardiography in the delineation of endocardial borders. Use of this technology

Figure 7. Demonstration of the development of harmonic frequencies due to nonlinear acoustic propagation through tissue.

should reduce the number of patients with ill- defined endocardial borders and result in a more accurate assessment of left ventricular f~nc t ion .~ Our preliminary experience suggests that tissue harmonic imaging also may result in superior visualization of intracardiac lesions such as clots (Fig. 6).

Harmonic imaging originally was developed for contrast agents to use the nonlinear nature of contrast microbubbles. Although it has been well known for many years that propagation through tissue does create some harmonic en- ergy, it has been assumed that the increased attenuation that occurs at the higher frequen- cies would attenuate the harmonics as quickly as they were produced. Thus, in the early days of harmonic contrast imaging, it was believed that the residual harmonic tissue image was due to incomplete rejection of the transmitted fundamental image. However, it was observed that the tissue harmonic image has some sig- nificantly improved characteristics that could not be explained by this description of this phe- nomenon. This effect is especially pronounced in technically difficult

Recent data suggest that the harmonic image is due to nonlinear propagation in tissue.8 The velocity of sound propagation depends on the density of the insonified material.g During the compression phase, the tissue becomes denser, and hence the ultrasound waves travel faster through the tissue than during the rarefaction phase; the compression phase tends to overtake the rarefaction phase (Fig. 7). This results in a distortion of the ultrasound waveform, which further develops with propagation distance.lO-" (Fig. 8). Due to these effects, the tissue tends to

Vol. 15, No. 5, 1998 ECHOCARDIOGRAPHY: A Jrnl. of CV Ultrasound & Allied Tech. 515

BECHER, ET AL.

Figure 8. Soundfield in conventional echocardiography (top) and tissue harmonic imaging (bottom). Rever- beration artifacts within the nearfield (e.g., ribs) appear as clutter artifacts in deeper areas. These artifacts con- tain only fundamental frequencies (blue). The harmonic components of the soundfield must be a certain dis- tance from the transducer to develop (red). If only harmonic components are processed, no reverberation arti- facts are displayed.

“generate” harmonics, shifting energy from the fundamental to the harmonic bands.

In technically difficult patients, there often is a diffuse haze caused by distortion of the trans- mitted beam by shallow surface layers or re- verberations between the skin surface and ribs. Because the transmitted beam has not yet had a chance to build up significant harmonic en- ergy, these distortions and reverberations are nearly entirely made up of ultrasound energy at the fundamental frequency. After the beam penetrates these shallow layers, it becomes more focused and begins to create the harmonic energy. However, once it has penetrated these layers, it is less affected by the more uniform tissue below these aberrating layers, and the harmonic beam retains its beam profile and fo- cusing characteristics. When the returned sig- nal is filtered at the harmonic, rejecting the fundamental, the clutter and haze are re- moved, leaving a much sharper image (Fig. 8). Another proposed reason for the diminished artifact and clutter with harmonic imaging is the reduction in the side lobe level in the sec- ond harmonic beam. It has been shown that harmonic imaging reduces side lobe level and main lobe width, leading to improved focusing

characteristics.10*’2 Although these effects are well established, it remains under investiga- tion whether they are the primary cause of the effects that have been described in this study.

Nonlinear effects in sound beam propagation are increased at higher ultrasound amplitudes. In this preliminary study, we did not systemat- ically study the effects of transmitted amplitude on harmonic imaging; however, the best results were obtained at high power levels, suggesting that the nonlinear effects of sound propagation are increased at higher emission power levels.

Limitations of the Study

The sequence of recordings in our study was not random because harmonic mode always followed the fundamental recordings. There- fore, some bias cannot be excluded because the investigator always knew which imaging tech- nique was active. In addition, in both imaging modes, the gain settings were not standard- ized. They were subjectively optimized by the examiner to obtain the best definition of endo- cardial borders. Also, the observers visually as- sessed the endocardia1 border definition score; this also is very subjective.

516 ECHOCARDIOGRAPHY: A Jrnl. of CV Ultrasound & Allied Tech. Vol. 15, No. 5,1998

TISSUE HARMONIC IMAGING

Conclusions

In conclusion, our preliminary study demon- strates the superiority of tissue harmonic imaging over fundamental B-scan imaging in the delineation of left ventricular endocardial borders in patients with suboptimal echocar- diograms. Harmonic imaging resulted in en- hancement and superior definition of endocar- dial borders and a decrease in noise and clut- ter artifacts.

1.

2.

3.

4.

5 .

References

Falcone RA, Marcovitz PA, Perez JE, et al: In- travenous albunex during dobutamine stress echocardiography: Enhanced localization of left ventricular endocardial borders. Am Heart J

de Jong N, Ten Cate FJ, Lancee CT, et al: Prin- ciples and recent developments in ultrasound contrast agents. Ultrasonics 1991;29:324-330. Burns PN: Harmonic imaging with ultrasound contrast agents. Clin Radio1 1996;51(suppl 1):

Tiemann K, Schlosser T, Likungu J, et al: Har- monic imaging without contrast: Does blood or tissue emit harmonic frequencies? (Abstract) Circulation 1997;96(suppl I):I-584. Erbel R: Echoventriculography: A simultane- ous analysis of two-dimensional echocardiogra-

1995;130:254-258.

50-55.

6.

7.

8.

9.

10.

11.

12.

phy and cineventriculography. Circulation

Schiller NB, Shah PK, Crawford M, for the American Society of Echocardiography Com- mittee on Standards, Subcommittee on Quan- tification of Two-dimensional Echocardio- grams: Recommendations for the quantifica- tion of two-dimensional echocardiography. J Am SOC Echocardiogr 1989;2:358-367. files JM, Sawada SG, Feigenbaum H, et al: En- hanced endocardial visualization using har- monic imaging without contrast. (Abstract) Cir- culation 1997;96(suppl I):I-585. Averkiou MA, Roundhill DN, Powers JE: A new imaging technique based on the nonlinear properties of tissues. IEEE Ultrasonics Sympo- sium, Toronto, Canada, 1997. Evans DH: Doppler Ultrasound: Physics, In- strumentation, and Clinical Applications. John Wiley 8z Sons, Chichester, 1989. Averkiou MA, Hamilton MF: Measurements of harmonic generation in a focused finite-ampli- tude sound beam. J Acoust SOC Am 1995;98:

Averkiou MA, Hamilton MF: Nonlinear distor- tion of short pulses radiated by plane and fo- cused circular pistons. J Acoust SOC Am 1997;

Christopher T: Finite amplitude distortion- based inhomogeneous pulse echo ultrasonic imaging. IEEE Trans Ultrason Ferroelec Freq Contr 1997;44:125-139.

1983;67: 205-210.

3439-3442.

102:2539-2548.

Vol. 15, No. 5,1998 ECHOCARDIOGRAPW: A Jrnl. of CV Ultrasound & Allied Tech. 517